Projects

Open Source

  • SyB3R allows to generate realistic images (including motion blur, noise, depth of field, etc.) along with ground truth for scene depth to benchmark SfM and MVS methods.
  • MVMS3D is a multi-view stereo tool that uses multiple shots per view point to increase the SNR and obtain a more complete 3D reconstruction. (The corresponding paper won the Best Paper by Young Authors Award at the ISPRS Congress 2016.)
  • StructuredRF aims to exploit densely labelled images for structured prediction.
  • SpeculaR implements an approach to detect and remove specular, shadow, or occluded regions in images of planar objects by replacing them with meaningful information from additional views.
  • GPU-BA provides GPU implementations of three different methods for bundle adjustment, namely the commonly used second order method based on Levenberg-Marquardt in two variations, as well as the two first order methods nonlinear conjugate gradients and alternating resection-intersection.
  • DropTrack implements a pipeline to detect and tracks drops in high-speed image sequences of a test cell developed at the Chair of Chemical and Process Engineering of Technische Universität Berlin.
  • UIPF is an open-source framework that allows to develop, implement, and visualize image processing tool chains.
  • VisualRF provides a Matlab script to render a given Random Forest Classifier as a set of visual binary trees in 3D.
  • FPCF implements an automatic, robust, and efficient fusion framework for point clouds ( in particular MS Kinect data ) based on coarse feature based alignment and fine ICP alignment.